Research On Vision Measurement Of Coal Flow Of Belt Conveyors Based On Deep Learning

Belt conveyor is an important equipment for coal mine production,with the characteristics of strong conveying capacity,long conveying distance,and continuous conveying.During the construction of smart mines,visual inspection devices are gradually used in coal mine production.Through the visual measurement of the coal flow carried by the belt conveyor,it can guide the optimization of motor speed adjustment,help to save energy,and ensure the safe operation of the main coal transportation system.In the process of binocular vision measurement of coal flow of belt conveyor,the repeated single color and texture of coal make the stereo matching of coal image fail.The uneven and random distribution of coal particles makes it difficult to calculate the coal packing density.This thesis studies the calculation method of coal depth information based on deep learning.The calculation method of coal packing density based on discrete element method and the distributed calculation method of coal flow of belt conveyor are studied in detail.The main work content is as follows:(1)A coal depth map calculation method based on deep learning is proposed,which includes coal image preprocessing and a coal disparity map calculation method based on the PSM-Net model.The coal image preprocessing part consists of three parts:image correction based on Bouguet algorithm,image segmentation based on Hough transform algorithm,and image enhancement based on histogram equalization.The PSM-Net model for coal image adopts pre-training and fine-tuning method to train.First,the Scene Flow and KITTI 2015 datasets are used to pre-train the model,and then the coal stereo matching datasets proposed in this thesis are used to fine-tune model.Compared with the traditional image stereo matching algorithm,the coal disparity map calculated by the method in this thesis can better describe the characteristics and the depth of the uneven distribution of the coal surface.(2)An online calculation method of coal packing density is proposed,which includes the calculation of the roundness and rectangularity of the coal packing surface image and the construction of an online calculation model for the coal packing rate.The CCD camera is used to obtain the surface image of the coal carried by the belt conveyor,and then perform morphological calculation,OTSU image segmentation,and edge calculation to obtain the roundness and rectangularity of the image of the coal packing surface.Then,according to the discrete element method,the online calculation method of coal packing rate is modeled to obtain the coal packing rate.Finally,according to the coal particle density,the coal packing density is calculated in real time.Numerical simulation shows that the average error of the proposed online calculation method for coal packing density is 2.349%.(3)An off-line calculation method of coal packing density is proposed,which includes the measurement of the length,width,and height of coal particles and the construction of an off-line calculation model for coal packing rate.By measuring the length,width,and height of a small number of coal pile particles,and then obtaining approximate surface area and volume of coal particles,the sphericity is calculated.Then,according to the discrete element method,an off-line calculation model of coal packing rate is constructed to obtain coal packing rate.Finally,according to the density of coal particles,the coal packing density is calculated.Numerical simulation shows that the average error of the proposed offline calculation method of coal packing density is3.389%.(4)A distributed method of calculating the coal flow carried by the belt conveyor is proposed,which includes the calculation of the volume of coal carried by the belt conveyor and the estimation of the distribution state of the coal carried by the belt conveyor.Calculate the difference between the volume of coal carried by the belt and the volume of the empty belt to obtain the volume of coal carried by the belt conveyor.Then,by obtaining the coal volume at the same location and at different times,the coal load distribution of the entire belt is obtained,and the coal flow of the belt conveyor is obtained.Compared with the traditional image stereo matching algorithm,the method proposed in this thesis has obvious advantages.The prototype system runs in 1127 ms and the coal flow measurement accuracy can reach 98.7043%.There are 51 figures,7 tables,and 97 references in this thesis.